This was a very well presented and insightful talk. One that I never knew I needed, but I enjoyed and will have to re-watch to increase the amount of information absorbed. Thank you Peter Bindels.
Saying that WAV files have a flat header is wrong, but the example still works. If you want to read a WAV file in general, you have to be mindful about a dozen different chunk types, chunks containing sub-chunks, and chunks being larger than usual (meaning they have software specific extra data). I know this because I'm currently reverse engineering a handful of WAV files that use a rare audio format and want to convert it to a more common one.
Really great introduction of this topic, situations where you need to reverse engineer existing files for valid reasons are actually quite often, so good job 👍
The networking world has things like ASN.1 which is effectively a grammar for data formats that allows you to generate encoders and decoders. Also several of the formats are like binary versions of text formats like xlm and json.
Can you please explain packaging versions with their versioned dependencies to google for Tensorflow and bazel build? (@1:08:00) I've spent dozens of hours try to get their ridiculous dependency trees set up just to get a cuda build with avx instructions - and not one or the other exclusively. It's mind-blowing stupidity to have such a sensitive system for everything from the compiler to the sdk's and system libraries, paths and configure files with hard coded launch commands of deprecated versions... Python 3.11 max, cudnn 8.6, bazel 3.1,0, tf 2.11.x, cuda 11 (not 12), clang 16, gcc 11, Linux kernels, pip, literally every single major component will break outside of a narrow range of versions and non more recent or current. Absolutely mind boggling from Google.
This was a very well presented and insightful talk. One that I never knew I needed, but I enjoyed and will have to re-watch to increase the amount of information absorbed. Thank you Peter Bindels.
Saying that WAV files have a flat header is wrong, but the example still works. If you want to read a WAV file in general, you have to be mindful about a dozen different chunk types, chunks containing sub-chunks, and chunks being larger than usual (meaning they have software specific extra data).
I know this because I'm currently reverse engineering a handful of WAV files that use a rare audio format and want to convert it to a more common one.
the best explanation of 2's compliment. that wrap around concept no one was taking about...
Great talk! More on reverse Engineering techniques please
Really great introduction of this topic, situations where you need to reverse engineer existing files for valid reasons are actually quite often, so good job 👍
Thanks
The networking world has things like ASN.1 which is effectively a grammar for data formats that allows you to generate encoders and decoders. Also several of the formats are like binary versions of text formats like xlm and json.
We’ll put together 👏
Can someone please recommend books that go in-depth through each step of the design and creation process for a complete noob 😅
Can you please explain packaging versions with their versioned dependencies to google for Tensorflow and bazel build? (@1:08:00) I've spent dozens of hours try to get their ridiculous dependency trees set up just to get a cuda build with avx instructions - and not one or the other exclusively.
It's mind-blowing stupidity to have such a sensitive system for everything from the compiler to the sdk's and system libraries, paths and configure files with hard coded launch commands of deprecated versions...
Python 3.11 max, cudnn 8.6, bazel 3.1,0, tf 2.11.x, cuda 11 (not 12), clang 16, gcc 11, Linux kernels, pip, literally every single major component will break outside of a narrow range of versions and non more recent or current. Absolutely mind boggling from Google.
This was amazing to watch, sadly the most interesting part was rushed and partially cut.
The reverse engineering part?
Inside a bit is the blood of a living AI.😂